fix(nimble): fix crash on disconnect due to stale per-connection CCCD pool
Closes BLERP-3057, BLERP-3060, BLERP-3061, and BLERP-3062
See merge request espressif/esp-idf!51868
test(wpa_supplicant): wait for the idle task after esp_wifi_deinit in eloop unit test
Closes IDFCI-11307 and IDFCI-12170
See merge request espressif/esp-idf!52111
The example enables CONFIG_ESP_SYSTEM_USE_FRAME_POINTER, which pushes the
bootloader to 0x6020 bytes on ESP32-H4. That exceeds the 0x6000 budget
available with the default partition table offset of 0x8000, breaking the
build for both the app_trace_jtag and app_trace_uart CI configurations.
Move the partition table to 0x10000 on ESP32-H4, same as already done for
ESP32-P4, and drop the now stale "only overflowed on P4" note. Tell OpenOCD
the matching app image offset so GDB can map flash on both chips.
- Update LC3 in btc and bta, update AT
- Update LC3 in SCO frame transfer and examples
- improve lc3 transfer speed
- Fix some encoding and decoding bugs
- Rename the functions and macros to fit both msbc and lc3
- Add config dependency
- Add migration guides for bluetooth-classic
- change the statement of "Wide Band Speech" to "Wideband Speech" to fit HFPv1.10
change(build): drop BUILD_COMPONENTS support and migrate test_apps to not use it for buildv2
Closes IDF-15859
See merge request espressif/esp-idf!51695
The Wi-Fi task deletes itself when esp_wifi_deinit() is called, and FreeRTOS
only reclaims its TCB and stack from the idle task afterwards. Test apps that
read the heap right after deinit therefore see that memory as still allocated
and report a leak, most visibly as the eloop unit tests failing on ESP32.
Wait for the idle task at every point where a test deinitialises Wi-Fi
before a leak check, replacing the single-tick delays that only matched what
esp_wifi_deinit() already waits for internally.
pytest.ini enables log_cli, so logging the full ninja stdout after a
failed first-time build is one multi-MB ERROR. That hung shard 3/6 for
hours after the hints test. Log the last 80 lines; keep the full output
on the exception and at DEBUG.
git worktree add materializes submodules as gitlink files. rmtree() cannot
remove those, and exists()+iterdir() on a source gitlink raises and falls
back to shutil.copytree, which leaves mbedtls/include as a file. Unlink
dest gitlinks and only copy populated source directories.
sdkconfig.ci.memprot_esp32s3 was built but listed in no CONFIGS_MEMPROT_*, so
the S3 PMS panic path was never exercised.
expect_gme() takes an optional core, for panics with no attributable core.
On Xtensa the PMS violation shares ETS_MEMACCESS_ERR_INUM with the cache error
and arrives as PANIC_RSN_CACHEERR. ESP32-S2 tells the two apart, ESP32-S3 never
did (//MV note in dd938eb95), so PMS faults were reported as "Cache error".
panic_memprot_fill_info() now claims the panic only when
esp_mprot_get_active_intr() reports a pending violation, and the ESP32-S3 cache
error path consults it, falling back to the cache error report as before.
RISC-V has a dedicated interrupt, so its reporting stays unconditional.
ESP32-S2 and ESP32 are untouched.
Move print_memprot_err_details() and its state out of the RISC-V panic_arch.c
into port/panic_memprot.c behind panic_memprot_fill_info(), so the report can
be shared with the Xtensa PMS targets. No functional change.
